A Probabilistic Template Update Method for Tracking Facial Tissue in Thermal Infrared

A novel template update method is proposed for facial tissue tracking in thermal clips. It is a smoothed matte approach, which provides the location along with the rate of updating. The template is able to adapt to abrupt orientation and physiological changes, while remaining robust to noise disturbances. Furthermore, it addresses successfully the difficult template drift problem. The new method was tested on tracking face regions in 40 thermal clips of individuals in varying psycho-physiological and environmental conditions. It demonstrated stability and accuracy, outperforming other template update strategies. The method promises improved performance in contact-free polygraphy and thus, it is of value in this emerging form of close range surveillance.